Split data folders into train and test for a GAN model in pytorch

I have two folders one of normal patients and another for abnormal patients and I would like to train the model just on health patient and test it on health and unhealth. How can I split the data ?

If u train it on just healthy patients, do u think it’ll predict accurately for unhealthy patients?

I am looking to detect the anomalies zones. So I think yes the model will be able to detect anomalies in the new images of unhealthy patient!